Every result shows how it was read.
We do not run a warehouse of pre-scraped pages. You point the API at a URL, we fetch it live, and the structured result carries a confidence score, the extraction method, and a fetch trail. Here is exactly where the data comes from.
Where the data comes from
| Source | How it is read | _provenance |
|---|---|---|
| The page you point us at | A /v1/scrape or /v1/extract call fetches the URL you pass, live, at request time. We do not resell a pre-scraped copy of the web. | json-ld · opengraph · regex · llm |
| Structured markup on that page | When a page publishes JSON-LD or Open Graph, we read those fields directly. The value is exact, not inferred, and costs no LLM credits. | json-ld · opengraph |
| Web search results | /v1/search returns ranked results from a search backend and, on request, scrapes each result the same way a direct scrape would. | search + fetch tier |
| Public records (coming to the public catalog) | Secretary-of-State and license-board lookups read official government registries. Those endpoints are not listed in the public developer catalog yet while we finish quality and documentation. When they run, facts and personal fields stay separate and tagged as public record. | us-public-record |
How we handle it
robots.txt respected by default
The default fetch honors a site’s robots.txt. Compliance is the default state, not a setting you have to find.
Live fetch, no silent cache
Every request is a live read unless you explicitly opt into a max-age cache. An agent acting on stale data acts wrong, so freshness is the default.
Provenance on every structured result
Structured results carry _confidence, _extraction_method, and _provenance (the fetch tier that won and how the result was read). You read how a value was obtained from the payload itself.
Takedown honored
A person or an owner can request removal of their record from the enrichment graph. Requests are processed through a documented takedown path.
Last updated 2026-07-15.
Read the provenance shape
The docs show the exact _confidence / _provenance envelope every structured result ships, and the takedown path for a record.